A Novel Loss Function for Deep Learning Based Daily Stock Trading System

Published: 11 November 2025| Version 2 | DOI: 10.17632/czwwfgcgz7.2
Contributor:
Ruoyu Guo

Description

This dataset contains all the data used in the paper "A Novel Loss Function for Deep Learning Based Daily Stock Trading System." The two files are for the 2004-2010 and 2018-2024 periods, respectively, containing most stocks in major US stock exchanges. Daily price (open, close, high, low) and volume (shares traded) data are downloaded from Yahoo Finance. Additional features are constructed from the price and volume data. We perform experiments on the two time periods to demonstrate our new proposed loss function is better at finding top growth opportunities in the stock market.

Files

Institutions

  • Brandeis University

Categories

Finance, Machine Learning, Convolutional Neural Network

Licence